Eleonora Bartoli

Motivated by the successes of large language models (LLMs), we hypothesized that semantic information in the brain may be expressed in distributed patterns of neural activity. We recorded responses of hippocampal neurons while individuals listened to narrative speech. Here, using an encoding model approach, we find robust encoding of semantics, even after controlling for phonemics and grammar. Individual words had complex selectivities that included multiple words in multiple semantic categories. Similar to embedding vectors in LLMs, distance between neural population responses correlated with semantic distance. For semantically similar words, even contextual embedders showed an inverse correlation between semantic and neural distances; we attribute this pattern to the noise-mitigating benefits of contrastive coding. Finally, neural population activity aligned most closely with GPT-2 embeddings, and variation in response patterns correlated with LLM-derived polysemy measures, indicating that semantic encoding is largely contextualized. Ultimately, these results provide a neurocomputational account for understanding how hippocampal neurons encode word meaning during language listening.

OBJECTIVE: Stereo-electroencephalography (sEEG) is a common method for clinical epilepsy monitoring and provides unique opportunities for intracranial research in humans. Optimal selection of reference electrodes is essential for obtaining high-quality localizable data. White matter (WM) electrode contacts are commonly used as references; however, this reference scheme presents several limitations that may influence the data, including a limited selection of electrodes in the WM, nonneutral activity in the WM, and a laterality bias. Here, the authors detail the use of a midline subgaleal (SG) electrode as an alternative reference for sEEG recordings. METHODS: An SG reference was used for 14 patients with drug-resistant epilepsy undergoing intracranial monitoring. Following the placement of sEEG electrodes, one 8-contact sEEG electrode (n = 2) or 4-contact strip electrode (n = 12) was placed in the SG space at the parietal midline. In a subset of 4 participants, we obtained awake, resting baseline recordings (5-minute duration) using different references, allowing us to compare signals recorded with an SG, WM, and gray matter (GM) reference. The authors compared the number of interictal spikes (IISs) detected by measuring the seizure onset zone selectivity index (SSI), cross-channel correlations, and power spectral density properties across these baseline recordings. RESULTS: No adverse effect of the SG electrode placement was reported in any participant. Recordings using an SG reference have a higher SSI compared with a WM or GM reference. Neural signals obtained with an SG reference have lower cross-channel correlations compared with the other two references and preserve more power at higher frequencies than a WM or GM reference. CONCLUSIONS: Extracranial placement of an SG electrode allows for a neutral midline reference. The authors’ findings demonstrate that an SG reference is a safe alternative to the standard WM reference, improving the signal-to-noise ratio and not interfering with the clinical investigation.

Intracranial neurophysiology studies have typically ignored signals from electrodes located in white matter (WM), assuming that their information content is artifactual or related to nearby gray matter (GM). Here, we tested the electrophysiological and functional features of signals recorded from different WM locations. Signals were recorded from 19 patients undergoing intracranial monitoring for drug-resistant epilepsy by means of stereo-electroencephalography (sEEG). Each sEEG electrode was classified into WM or GM based on the surrounding tissue. We obtained recordings from a total of 1,717 sEEG electrode contacts, 36% in WM, while the patients were in awake resting state (5 minutes). For each sEEG electrode, we employed a model-based spectral decomposition to separate periodic and aperiodic components, and we computed signal complexity metrics. For a subset of participants, we computed WM structural information from diffusion-weighted magnetic resonance imaging and we evaluated functional signals during a cognitive control task. Our results show that signals recorded from WM have different spectral features and higher complexity than GM. Complexity correlates positively with fractional anisotropy, and modulations related to behavior during the task were detected in WM. Overall, this indicates that WM signals carry information that may reflect signal propagation across WM fiber tracts.

Finding suitable therapies for treatment-refractory neuropsychiatric disorders constitutes a major goal for translational neuroscience. Deep brain stimulation shows promise for treatment resistant depression, but treatment efficacy varies substantially across patients. Objective, electrophysiologically driven strategies to optimize deep brain stimulation for treatment resistant depression could greatly improve clinical efficacy by minimizing the trial-and-error approach needed to personalize stimulation settings. This may not only reduce the delay between the start of the treatment and symptom improvement, but also enable acute, real-time verification of circuit engagement, advancing our understanding of the mechanism mediating antidepressant effects. Here, we investigate whether cerebro-cerebral evoked potentials elicited through different deep brain stimulation configurations could be used to guide stimulation personalization for treatment resistant depression. Cerebro-cerebral evoked potentials offer a fast, objective way to identify regions engaged by stimulation, revealing the effective connectivity pattern of the stimulated location. Data were collected from eight patients with treatment resistant depression who received dual bilateral deep brain stimulation devices targeting the subcallosal cingulate and ventral capsule/ventral striatum. During an initial in-hospital monitoring period, single-pulse electrical stimulation was delivered through the deep brain stimulation devices and cerebro-cerebral evoked potentials were recorded through temporary stereo-electroencephalography probes across fronto-temporal regions. Patients underwent several outpatient stimulation programming sessions over the course of 9 months to identify the stimulation configurations leading to the greatest improvement in depressive symptoms. We retrospectively analysed cerebro-cerebral evoked potentials obtained in response to stimulation of different stimulation configurations to identify features distinguishing the clinically effective configurations. The deep brain stimulation configurations leading to the greatest improvement in depressive symptoms were associated with significantly larger evoked potentials in the orbitofrontal cortex and showed an increased number of evoked potentials across dorsal and ventral prefrontal regions. Waveform similarity analysis revealed a gradient in therapeutic effects, such that multiple alternative stimulation configurations led to similar symptom improvement. The vast deep brain stimulation parameter space might contain a configuration subspace defined by comparable therapeutic effects. In addition, evoked potentials obtained from single-pulse and from bursts of high-frequency stimulation displayed similar spatial patterns, suggesting that either method might be able to identify the configuration best engaging the circuit mediating the clinical response. Together, these findings provide proof-of-principle evidence that stimulation-evoked prefrontal responses reflect network engagement associated with antidepressant effects. Cerebro-cerebral evoked potentials may offer an objective and acute strategy to guide contact selection in deep brain stimulation for treatment resistant depression.

Neural oscillations play a critical role in shaping neuronal firing patterns. While phase-locked neuronal firing (“phase tuning”) has been extensively studied in animal models and human invasive recordings, much less is known about whether neurons show preferential firing at specific oscillatory frequencies, termed frequency tuning. Here, we employ human intracranial recordings across several brain regions including hippocampus, entorhinal cortex, anterior and posterior cingulate cortex, and orbitofrontal cortex to test the hypothesis that neurons exhibit frequency-specific firing. We analyzed 357 single units recorded simultaneously with local field potentials in 19 neurosurgical patients during awake resting. We estimated the instantaneous frequency of the LFP using adaptive spectral decomposition and assessed frequency tuning of each neuron while controlling for changes in firing rate unrelated to frequency changes. We found 27% of neurons exhibited increased or decreased firing within specific frequencies, most commonly within the low-frequency range (<10 Hz). Neurons exhibiting frequency tuning were distinct from those displaying phase tuning, and both types of tuning were observed across multiple brain regions with no anatomical preference. Together, our results demonstrate that the instantaneous frequency of neural oscillations modulates neuronal firing which may serve as an additional mechanism for information processing in the human brain, opening new avenues for frequency-targeted neural stimulation.

Background: Invasive neuromodulation therapies such as deep brain stimulation (DBS) have become increasingly important for treating refractory neurological and psychiatric disorders, including movement disorders, epilepsy, obsessive-compulsive disorder, and more. Despite increasing clinical use, optimizing brain stimulation remains a major challenge. Key questions remain regarding how to select optimal stimulation targets, determine electrode placement, program the large parameter space of stimulation settings, understand mechanisms of action, and adapt therapy as symptoms fluctuate or progress over time. Because direct clinical symptom measurement can be variable, delayed, or difficult to quantify intraoperatively, physiologic biomarkers hold potential to navigate these challenges. One promising approach is to measure the neural responses elicited by brief pulses of electrical stimulation, based on the assumption that stimulation-evoked potentials reflect the complex connectivity of activated brain networks. Objective: Following PRISMA guidelines, we systematically review studies investigating how such stimulation-evoked potentials, particularly cerebro-cerebral evoked potentials (CCEPs), can improve invasive brain stimulation in humans. Results: Across movement disorders, epilepsy, and psychiatry, stimulation-evoked potentials have been used to determine network engagement, refine target localization, guide postoperative device programming and, in emerging cases, inform adaptive or state-dependent stimulation strategies. Conclusion: We highlight common methodological approaches, key findings, and limitations across these applications. Finally, we discuss future directions needed to transition from retrospective circuit characterization toward prospective, patient-specific guidance of brain stimulation therapies.

A hallmark of human cognitive flexibility is the ability to perform a wide range of unrelated tasks. While much is known about how neural circuits support individual tasks, less is understood about how these circuits support multiple tasks. Are neural representations largely task-specific, or do they retain a shared structure across distinct behaviors? To investigate this question, we recorded neural population activity from the hippocampus and anterior cingulate cortex, two regions linked to generalization and cognitive control, in eleven human patients performing three distinct tasks. Using dimensionality reduction, we estimated the neural subspace associated with each task and compared subspace geometry across tasks. We found that task-related subspaces were not independent: across tasks, approximately half of the subspace dimensions were shared, and these were primarily the dimensions containing most of the shared neural covariance. These findings indicate that neural population activity in these regions is not purely task-specific. This suggests a stable, low-dimensional circuit structure may persist even across unrelated behaviors, potentially providing a common substrate for flexible cognition.